Pinokio
1

🚀 Advanced Version Selection for ComfyUI Installer

@maoperposted 5/23/2026, 10:18:11 PM·Owner·0 replies

🚀 Advanced Version Selection for Inteliweb ComfyUI Installer

The Inteliweb ComfyUI Installer for Pinokio now gives users more control over their ComfyUI installation by allowing them to select specific versions of:

  • ComfyUI
  • Python
  • PyTorch / CUDA

This is especially useful when working with advanced workflows, custom nodes, or AI models that require specific software versions for proper compatibility.

Repository:
https://github.com/maoper11/inteliweb-comfyui

Pinokio App:
https://beta.pinokio.co/apps/github-com-maoper11-inteliweb-comfyui

Why version selection matters

ComfyUI is constantly evolving, and many custom nodes are updated frequently. However, not every workflow or node works perfectly with the latest version of every dependency.

Some workflows may require a specific combination of:

  • ComfyUI version
  • Python version
  • PyTorch version
  • CUDA version
  • TorchVision / TorchAudio
  • Custom compiled wheels
  • Optimization libraries such as Flash Attention, Triton, SageAttention, NATTEN, xFormers, or other dependencies

Because of this, installing only the latest version is not always the best option.

Sometimes, the most stable setup is the one that matches the exact environment used by a tutorial, workflow developer, or custom node package.

Available configuration options

COMFY_VER

COMFY_VER allows you to install either the latest ComfyUI version or a specific fixed version.

Examples:

COMFY_VER=latest
COMFY_VER=v0.22.0
COMFY_VER=v0.21.1

Recommended for most users:

COMFY_VER=latest

Use latest if you want the newest available ComfyUI version.

Use a fixed version if:

  • A workflow was tested with a specific ComfyUI release
  • A custom node is not yet compatible with the latest version
  • You want to reproduce the exact environment used in a tutorial
  • You need a more stable installation for production work

PYTHON_VER

PYTHON_VER allows you to select the Python version used to create the ComfyUI environment.

Examples:

PYTHON_VER=3.12
PYTHON_VER=3.11
PYTHON_VER=3.10

Recommended for most users:

PYTHON_VER=3.12

Python 3.12 is recommended for most modern ComfyUI installations.

However, some older custom nodes may work better with Python 3.10 or 3.11, especially when they depend on libraries that have not yet been updated for newer Python versions.

TORCH_VARIANT

TORCH_VARIANT allows you to select the PyTorch and CUDA configuration according to your GPU, operating system, and workflow requirements.

Examples:

TORCH_VARIANT=auto
TORCH_VARIANT=latest-cu130
TORCH_VARIANT=2.10.0-cu130
TORCH_VARIANT=2.9.1-cu128
TORCH_VARIANT=2.8.0-cu128
TORCH_VARIANT=directml
TORCH_VARIANT=2.7.0-rocm6.3
TORCH_VARIANT=cpu

Recommended for most users:

TORCH_VARIANT=auto

When using auto, the installer will try to select the best PyTorch configuration for your system automatically.

Advanced users can manually select a specific Torch variant when they need compatibility with certain GPUs, CUDA versions, custom nodes, or precompiled wheels.

Recommended setup for most users

For a simple and stable installation, use:

COMFY_VER=latest
PYTHON_VER=3.12
TORCH_VARIANT=auto

This is the easiest option and is recommended for most users.

The installer will automatically detect your system and choose the most appropriate configuration.

Advanced NVIDIA setup

For advanced NVIDIA users who want to test a newer PyTorch and CUDA environment:

COMFY_VER=latest
PYTHON_VER=3.12
TORCH_VARIANT=latest-cu130

This can be useful for users working with newer GPUs, recent CUDA builds, or workflows that depend on newer PyTorch features.

Compatibility setup for older workflows

For workflows or custom nodes that require a more specific environment:

COMFY_VER=v0.21.1
PYTHON_VER=3.12
TORCH_VARIANT=2.8.0-cu128

This helps avoid errors caused by recent changes in ComfyUI, PyTorch, CUDA, or custom node dependencies.

When should you use a specific version?

Use specific versions when:

  • A workflow was tested with an exact ComfyUI release
  • A custom node does not work correctly with the latest ComfyUI version
  • A special wheel requires a specific PyTorch / CUDA version
  • You are following a tutorial that uses a fixed environment
  • You need better stability for production work
  • You want to reproduce the same installation across multiple computers
  • You are troubleshooting dependency conflicts
  • You are working with advanced nodes that require specific compiled libraries

Main advantages

✅ One-click ComfyUI installation through Pinokio
✅ Manual ComfyUI version selection
✅ Manual Python version selection
✅ Manual PyTorch / CUDA version selection
✅ Automatic GPU detection
✅ NVIDIA, AMD, macOS, and CPU support
✅ Better compatibility with advanced workflows
✅ Easier reproduction of tutorial environments
✅ More control for creators, developers, and ComfyUI power users

This update makes the Inteliweb ComfyUI Installer much more flexible.

Beginners can keep the default automatic setup, while advanced users can manually select the exact versions required by their workflows, custom nodes, models, or tutorials.

Built with ❤️ by Inteliweb AI

Replies (0)
Up to 10 files, 25MB each. Images are optimized; GIFs -> MP4; videos 720p (max 120s).
🚀 Advanced Version Selection for ComfyUI Installer · Pinokio